Path Analysis Modeling Indicates Free Transport Increases Ambulance Use for Minor Indications

Path Analysis Modeling Indicates Free Transport Increases Ambulance Use for Minor Indications Joseph Yuk Sang Ting, MBBS, B Med Sci 1, 2 and Allan M. Z. Chang, MBBS, PhD 3 1 Department of Emergency Medicine, Mater Public Adult Hospital, South Brisbane, Australia. 2 Division of Anesthesiology and Critical Care, School of Medicine, Southern Clinical School, Faculty of Health Sciences, University of Queensland, Brisbane, Australia. 3 Mater Research Support Center, Mater Health Services, South Brisbane, Australia. ABSTRACT Introduction. Clinically unnecessary ambulance transport is increasing, diverting limited resources from patients needing ambulance transport. It was anecdotally observed that inappropriate ambulance use increased after abolition of a direct patient cost for ambulance transport. Hypothesis. In July 2003, direct patient fees were abolished in favor of a universally applied ambulance levy, potentially leading to increased ambulance use by patients with low illness acuity and admission rates. Methods. The influence of age, illness acuity, and need for admission on ambulance use was assessed for 55,397 emergency department attendances in 2002 and 2004. Ambulance users were compared with nonusers in both years and attendances for 2002 compared with 2004 using chi-square test for two groups. Logistic regression provided a multivariate model leading to ambulance use. Path analysis modeling to assess interrelationships between factors associated with ambulance use was developed. Results. Ambulance users in both years were older, had more acute illness, and had greater need for admission compared with nonusers. The odds ratio (OR) of arrival by ambulance in 2004 compared with 2002 was 1.14 (95% confidence interval, [CI], 1.12 to 1.17). In 2002, ambulance users were older (OR, 1.42; 95% CI, 1.40 to 1.43), were more likely to need admission (OR, 2.28; 95% CI, 2.16 to 2.4) and had higher illness acuity (OR, 2.02; 95% CI, 1.94 to 2.09). There was a negative correlation between 2004 and illness acuity. Conclusions. Ambulance use increased in 2004 after patient transport fees were abolished. Increased use was associated with decreased age, clinical acuity, and admission need. Abolishing direct patient cost stimulates ambulance use, potentially including inappropriate transport. Path analysis to assess the effect of changed funding on ambulance use could be used to the influence of other locally relevant factors contributing to ambulance use. Keywords: ambulance use; path analysis; free patient transport. INTRODUCTION Understanding who, when, and for what an ambulance is used will help in developing a monitoring mechanism to reduce clinically unnecessary patient transport, thereby improving ambulance

utilization.[1] As far as we are aware, no surveillance mechanism has been developed to monitor the relative and interrelated contributions of clinical and demographic factors influencing ambulance use. Most ambulances are used to transport patients to the emergency department (ED) for assessment and treatment. The intersection between patients arriving by ambulance and the ED therefore provides a representative situation to examine the patterns of ambulance use by the public.[1, 2] A change in funding structure from direct patient cost to a universally applied ambulance levy in July 2003 resulted in anecdotal observations by ED and prehospital staff that more patients were being transported for minor conditions. It was surmised that ambulances that incurred no direct patient cost would be more prone to being inappropriately used, as reflected in the lesser acuity and admission need of those transported. To study this hypothesis, a path analysis model to quantify the contribution of patient clinical and demographic factors leading to request for an ambulance was developed. Clinically inappropriate ambulance use is a serious and widespread problem. Clinically unnecessary ambulance use varies widely in different localities, making comparisons difficult.[3, 4] In a retrospective review of 300 consecutive ambulance transports by emergency physicians in London, 15.7% were considered inappropriate,[5] while 36% of ambulance transports in Dublin were deemed inappropriate by emergency physicians.[6] In the United States, inappropriate ambulance use occurs in up to 47% of adult cases,[4] being influenced by health insurance and socioeconomic status.[4, 7] Inappropriate ambulance use increases response times to clinically urgent cases by diverting limited ambulance resources to nonurgent cases,[1, 8, 9] contributes to unsustainable demand,[1] is expensive[9] and lowers ambulance staff morale.[8] Demand for ambulance services, including that for inappropriate transport,[4] is increasing rapidly.[1] It is therefore important that ambulances are utilized by patients only when a clinical need exists.[10] Although ambulances should be provided primarily on the basis of clinical need,[11] there are no consistent prehospital[3, 12, 13] or hospital[3, 5, 7, 13] criteria defining clinical need and appropriate use[5, 8, 14] other than that for patients requiring resuscitation and those with suspected myocardial infarction, stroke, and other time-critical conditions.[15, 16] The goals of this study were 1) to determine whether an effectively free ambulance service incentivizes its use by patients with lower acuity and admission rates and 2) to develop a method for monitoring ambulance activity to achieve the first goal. The model should be capable of quantifying the influence of, and examining the interactions between, factors believed to affect ambulance use. METHODS Setting The Mater Health Services Research Ethics Committee determined that this was a quality assurance study and was therefore exempt from independent ethical review as specified by the National Health and Medical Research Council. The Mater Adult Public Hospital, an urban tertiary limited-trauma ED within 5 km of the central business district of Brisbane, provides free health care for adults and has an annual attendance of 28,000. The Mater Adult Public Hospital is within close proximity of a predominantly inner-city population, comprising high-density living, established older inner-city suburbs, and underprivileged homeless residents. It is the sole tertiary care provider for complications of early pregnancy for a catchment population of 750,000 in greater south Brisbane. One major trauma center and several peripheral public (free) hospitals serve greater south Brisbane. The Queensland Ambulance Service provides a comprehensive multitiered emergency transport and prehospital response service staffed by paramedics trained to different skill levels, the highest of which are airway, thoracostomy, and vascular access skills. Before July 2003, one third of Queenslanders were pensioners or their dependents covered under the State Government s Free Ambulance Transport for Pensioners policy, one third were ambulance subscribers, and the remaining one third possessed private health insurance with an ambulance component or had no coverage at all. Only those with no coverage (a portion of the last group) paid a fee for transport, which depended on urgency and distance to hospital. As such, more than two thirds of Queenslanders (pensioners, ambulance subscribers, and privately insured patients) were transported without incurring a fee before July 2003. After July 2003, all Queenslanders were transported at no direct cost with the introduction of a universally applied levy, the Community Ambulance Cover. Any change in ambulance use between 2002 and 2004 may be partly attributable to the abolition of fee for transport.

Data Collection Data for ED attendances in 2002 and 2004 were extracted using focused search algorithms from Emergency Department Information Systems, a computerized ED time-patient flow management system incorporating patient details, triage complaint, triage score, diagnostic tests, final diagnosis, and disposition. Measurements Patient age, time of arrival, and postcode were edited and categorized so that they more closely reflect parameters associated with ambulance use and are more compatible with analysis of frequencies. Personal data include age (categorized in decades) and gender. Environmental data include distance from hospital (postcodes were edited to within or greater than 10 km of hospital), time of arrival (edited to day [8 AM to 8 PM] or night [8 PM to 8 AM]), day of arrival (edited to weekday [Monday to Friday] or weekend [Saturday and Sunday]), and year of arrival (2002 or 2004). Clinical data include acuity, which is the reverse of the National Triage Score (NTS) (acuity = 6 NTS), whether the patient was discharged or admitted after ED assessment, and the primary diagnosis [28 aggregated International Classification of Diseases-10 diagnostic codes[17]). The NTS is a score from 1 to 5 of clinical urgency that determines the order of assessment by ED medical staff. A lower NTS requires more urgent assessment and treatment. The NTS has been validated in terms of clinical urgency, admission rates, and hospital outcome and has good interobserver and interinstitutional reliability.[17] It is widely used in Australasia and is one of the Australian Council on Health Care Standards Clinical Indicators for Emergency Medicine.[18] As the NTS signifies the priority for treatment, the term acuity (acuity = 6 NTS) reverses the scale so that a high-acuity score represents a lower NTS, with higher-acuity conditions requiring more urgent medical attention. Statistical Descriptions The Data A primary analysis was unable to demonstrate a relationship between aggregated diagnostic category and other parameters. A substantial proportion of cases did not have a diagnosis entry (3,184; 5.7% of all attendances), and the diagnostic categories were based on broad disease groupings that did not have an identifiable relationship with acuity of illness or with type of transport used. The categoryaggregated diagnostic groupings were therefore excluded from multivariate analysis. Statistical Method Because the study sample size was large and variables were easily categorized, demographic and other data were analyzed using frequency tables. Primary bivariate comparisons were performed using the chi-square test for two groups [19]: comparing data between 2002 and 2004 and comparing ambulance users with nonusers in both years. Logistic regression [20] was used to provide a multivariate model for ambulance use. Logistic regression is a rigorous and robust method to build multivariate models and for hypothesis testing. This is therefore the main method where factors believed to influence ambulance use are statistically tested for significance. Path analysis modeling, [21] incorporating linear regression coefficients, was used to examine the cascade of sequential events that led to ambulance use. This method allows partitioning of variances and permits the assumption that early variables sequentially affect those downstream in the cascade. Path analysis enables examination of complex interrelationships between variables that lead to request for an ambulance. The model uses three levels in a defined sequence. Personal and environmental factors influence acuity and need for admission, and together with them influence the use of ambulance.

Because acuity and need for admission are determined after patient arrival in the ED, they are used as surrogate measures of illness severity at the time of ambulance request. The model evaluates the strength of these interrelated influences on ambulance use. RESULTS Data from 55,397 ED attendances in 2002 and 2004 were compared, and the results are presented in Table 1. Patients arriving in 2004 had lower-acuity illness and were less likely to require admission. A higher proportion of patients in 2004 were 30 50 years of age and living within 10 km of the hospital. These patients would be expected to be more capable of coming to the hospital independently, yet they used the ambulance more. TABLE 1. Comparison of Data Between 2002 and 2004, Including That Between Ambulance Users and Those Who Did Not Use the Ambulance 2002 2004 Chi-square Power (1 ß) Day Weekday 20,116 (72.6) 19,872 (71.8) Weekend 7,611 (27.4) 7,798 (28.2) 3.66 (NS) 0.49 Time Day 19,706 (71.1) 19,373 (70.0) Night 8,021 (28.9) 8,297 (30.0) 7.39* > 0.99 Gender Female 15,514 (56.0) 15,554 (56.2) Male 12,211 (44.0) 12,111 (43.8) 0.39 (NS) 0.09 Age in decades 0 6 (0.0) 0 (0.0) 1 2,552 (9.2) 2,300 (8.3) 2 7,069 (25.5) 6,925 (25.0) 3 5,510 (19.9) 5,882 (21.3) 4 3,717 (13.4) 3,828 (13.8) 5 2,860 (10.3) 2,909 (10.5) 6 2,043 (7.4) 2,054 (7.4) 7 2,079 (7.5) 2,036 (7.4) 8 1,527 (5.5) 1,392 (5.0) 9 350 (1.3) 329 (1.2) 10 10 (0.0) 13 (0.0) 36.49 0.32 Distance < 10 km 13,403 (48.3) 14,098 (51.0) > 10 Km 14,324 (51.7) 13,572 (49.0) 37.67 > 0.99 Acuity (6 NTS %) 1 (6 5%) 1,907 (6.9) 2,324 (8.4) 2 (6 4%) 16,258 (58.6) 16,900 (61.1) 3 (6 3%) 8,074 (29.1) 7,239 (26.2) 4 (6 2%) 1,414 (5.1) 1,150 (4.2) 5 (6 1%) 74 (0.3) 57 (0.2) 128.39 > 0.99 Disposition Discharged 21,142 (76.3) 21,305 (77.0) Admitted 6,585 (23.7) 6,365 (23.0) 4.26 0.54 Transport Self 22,282 (81.2) 21,655 (78.5) Ambulance 5,151 (18.8) 5,927 (21.5) 62.73 > 0.99 NS = not significant; NTS = National Triage Score, rated 1 to 5. *p < 0.01. p < 0.001. p < 0.005.

Logistic regression coefficients were determined for the likelihood of ambulance use as it relates to age, acuity, and need for admission, among others. These are presented in Table 2 as adjusted odds ratios with 95% confidence intervals. All variables are assumed to have simultaneously influenced ambulance use. Ambulance users were significantly more likely to need admission, have an acute illness, arrive at night, belong to an older age group, live more than 10 km from the hospital, be female, and arrive in 2004. TABLE 2. Logistic Regression Analysis Modeling the Relationship Between Personal, Environmental, and Clinical Parameters and Ambulance Use Predictor Variable Adjusted Odds Ratio 95% Confidence Interval Correlation Coefficient Standard Error Z p- value Nighttime* 1.5295 1.4540 1.6091 0.425 0.0259 16.4359 < 0.001 Older age group 1.4175 1.4013 1.4340 0.3489 0.0059 59.3342 < 0.001 Needs admission 2.2793 2.1619 2.4031 0.8239 0.027 30.5411 < 0.001 > 10 km from 1.1490 1.0957 1.2049 0.1389 0.0242 5.7312 < 0.001 hospital Female 1.1305 1.0778 1.1855 0.1226 0.0243 5.0427 < 0.001 Acuity (6 NTS) 2.0198 1.9456 2.0934 0.7029 0.0183 38.3786 < 0.001 Weekend 1.0419 0.9884 1.0982 0.041 0.0269 1.5267 0.1268 Year 2004 1.1455 1.1188 1.1729 0.1359 0.012 11.2764 < 0.001 Constant * * * 274.187 24.1296 11.3631 < 0.001 *Day-8 AM to 8 PM; night = 8 PM to 8 AM. NTS = National Triage Score, rated 1 to 5. The results of path analysis are presented in Figure 1, demonstrating only coefficients that are significant to p < 0.001. All variables other than day of arrival have a statistically significant relationship with ambulance use (day of arrival was therefore excluded from the model). Ambulances were used more in both years by older patients (0.27), those who had more acute illness (0.17), those needing admission (0.16), those arriving at the ED at night (0.06), those living more than 10 km from the hospital (0.03), and females patients (0.02). Increased ambulance use in 2004 (0.04) was, however, associated with less acute illness ( 0.05) and living closer to the hospital ( 0.03). Correlations within each level exist within the path analysis model. Acuity and need for admission are related (0.36) because both reflect illness severity. Secondary influences reflecting interactions were also demonstrated. For instance, female patients (0.02), older patients (0.16), and those arriving at night (0.09) had more acute illnesses and required admission, with these in turn leading to increased ambulance use. DISCUSSION In our study, free ambulance transport was associated with increased clinically inappropriate transport as indicated by declining illness acuity, patient admission, and younger patient age, with the exception of young adult multitrauma victims. The latter issue was not specifically examined in the study. To our knowledge, this is the first attempt to develop an ambulance use path analysis model capable of demonstrating independent changes in clinical acuity, admission rate, patient age, and any other chosen factor(s) among ambulance users. The quantitative surveillance over time at our ED of these three clinical, demographic, and attendance characteristics, among others, led to the detection of increased nonurgent ambulance use following abolition of an ambulance transport fee. Previous studies[5, 6, 8, 12] have only measured rates of clinically inappropriate ambulance use as determined retrospectively by emergency clinicians.

FIGURE 1. Path analysis showing cascading and quantified influence of factors leading to ambulance use. More than two thirds of the community (ambulance subscribers, pensioners and their dependents, privately insured) were not affected by the introduction of the ambulance levy in July 2003 because they were already transported at no personal cost. Significantly increased ambulance use in 2004 would reasonably be assumed to have occurred in the one third of the community who, having previously been charged a fee for transport, now had this fee waived. This occurred despite decreased acuity and admission rates among ambulance users in 2004. Furthermore, younger patients and those living closer to the hospital used an ambulance more frequently in 2004. It is reasonable to surmise that when users do not incur fees for transport, they will use ambulances for less acute illnesses despite being more capable of getting to the hospital without assistance. On the other hand, charging transport fees and restricting eligibility risks ambulances not being used by clinically acute patients who cannot afford to pay a fee.[5, 8] The decline from 2002 to 2004 in this study of clinical acuity, admission rate, and age of ambulance users may represent an increase in unnecessary ambulance use. Although not seen in this study, such a trend would prompt correction by health system or public education interventions, such as improved ambulance dispatching triage systems and scene treatment for minor illnesses.[6, 8] Scarce ambulance resources could then be redirected to patients with greater clinical need for ambulance treatment and transport, that is, patients with high illness acuity, patients with need for admission, and older patients. Aside from being sicker, these patients are less mobile.[11] Additional factors, such as access to community health care, nonambulance private and public transport, and socioeconomic status, could be incorporated into the model, with both their individual and interrelated influence on ambulance use evaluated. As such, path analysis modeling may be useful for planning and improving ambulance services and as a tool for further prehospital and health systems research. Among ambulance users in both years and in decreasing magnitude of influence, older age, acuity, need for admission, nighttime arrival, and living more than 10 km from the hospital were each independently associated with increased ambulance use. Previous studies have demonstrated an association between older age, higher acuity, and admission rates with ambulance use.[6, 7, 22] Age older than 65 years, higher acuity, and admission to the hospital predicted ambulance use in a study of 10,229 patients transported to a Queensland hospital in 1996 1997.[2] Older patients place high demands on ambulance services, EDs, and hospital beds.[2, 11, 23] The elderly have more acute illnesses, are more likely to be admitted,[23] and have reduced ability to use, and access to, non-ambulance transport.[2, 11] The combination of these factors with decreased physical and cognitive function in the elderly[8] leads to greater but more appropriate ambulance use in the elderly.[4, 6, 7]

As in this study, ambulance arrivals were more likely at night to a Montreal ED.[22] Reduced access to non-ambulance transport and primary health care at night may explain this finding. However, unlike in this study, time of arrival was found to not affect ambulance use in Ipswich Hospital,[2] 45 minutes by car from our hospital. This reflects wide variation in even local patterns of ambulance use. LIMITATIONS Increased ambulance use for less acute conditions in 2004 compared with 2002 was adjusted for other pre-specified variables, was of a large magnitude, and achieved significant p-values of at least < 0.05. It is therefore unlikely that this change in ambulance use could be attributed to secular differences in disease presentation and population patterns, longer-term trends in the use of emergency services, and changes in health systems. This could be clarified by tracking ambulance activity over a longer period. Six-month periods before and after the introduction of the ambulance levy were excluded to reduce transitional effects of funding change. There are no reliable definitions or measures of appropriate ambulance use by a patient.[5, 8, 14] Broad criteria include whether treatment was potentially required en route to the hospital,[13] whether rapid transport to hospital was needed[3] and the reason for the patient requesting an ambulance.[12] Other studies use individual patient-based criteria such as ED assessment[7, 12] or diagnosis,[14] hospital diagnosis, and patient disposition to retrospectively determine whether transport was necessary.[6] This study focused on factors associated with ambulance use by an overall patient population rather than any individual. Because ambulances are requested by the patient (consumer controlled), the need for an ambulance as perceived by the patient invariably exists. A patient s perception of need frequently differs from medical opinion.[5, 8, 14] Our model cannot determine whether appropriate use of an ambulance has occurred from the patient s point of view. A patient s decision to call an ambulance for transport to the ED is influenced by personal circumstances and the health care resources the patient is able to access. Each patient has specific reasons for using an ambulance, and whether ambulance use is justified on that occasion for that patient is difficult to determine. These reasons include heightened perception of urgency, lack of access to alternative transport and primary care, the belief that arrival at the hospital by ambulance will hasten time to being seen, and, crucially, the personal financial cost of using an ambulance.[2, 5] Because this model is developed from ED attendances, it is unable to evaluate the effect on ambulance use of non hospital-based health resources such as primary health care facilities and afterhour general practice services, although this could be assessed in future models. CONCLUSIONS Ambulance use increased in 2004 after direct patient fees were abolished in favor of a universally applied ambulance levy. Increased ambulance use was associated with reductions in patient age, clinical acuity of conditions and need for admission. A change in ambulance funding, such as the removal of direct patient cost, stimulates ambulance use, potentially including inappropriate requests for transport. The path analysis model developed to study the effects on ambulance use of changes in ambulance funding structure may be used to determine influence of other locally relevant factors contributing to ambulance use. This may be conducted at one health care facility over time or comparison made between facilities at one point in time. Modeling the dynamics of ambulance use facilitates the assessment of appropriate ambulance use and the effect of interventions to improve ambulance utilization. ACKNOWLEDGEMENT The authors thank Dr. K. Humphrey for data abstraction and Dr. C. Foot, Dr. M.Y. Ling, and Dr. E. Merfield for their thoughtful appraisal of the manuscript. REFERENCES 1. Australian Institute for Primary Care Ambulance demand and funding project Faculty of Health Sciences, La Trobe University: Melbourne, Australia, (2004).

2. Clark, M J, Purdie, J, FitzGerald, G J, Bischoff, N G, and O Rourke, P K Predictors of demand for emergency prehospital care: an Australian study. Prehospital Disaster Med (1999), 14: 167 72. 3. Hauswald, M, and Jambrosic, M Denial of ambulance transport: can reviewers determine what is an emergency?. Prehosp Emerg Care (2004), 8: 162 5. 4. Richards, J R, and Ferrall, S J Inappropriate use of emergency medical services transport: comparison of provider and patient perspectives. Acad Emerg Med (1999), 6: 14 20. 5. Palazzo, F F, Warner, O J, Harron, M, and Sadana, A Misuse of London ambulance service: how much and why?. J Accid Emerg Med (1998), 15(6): 368 70. 6. Little, G F, and Barton, D Inappropriate use of the ambulance service. Eur J Emerg Med (1998), 5: 307 11. 7. Rucker, D W, Edwards, R A, Burstin, H R, O Neil, A C, and Brennan, T A Patient-specific predictors of ambulance use. Ann Emerg Med (1997), 29: 484 91. 8. Pennycook, A G, Makower, R M, and Morrison, W G Use of the emergency ambulance service to an inner city accident and emergency department a comparison of general practitioner and 999 calls. J R Soc Med (1991), 84: 726 7. 9. Marks, P J, Daniel, T D, Afolabi, O, Spiers, G, and Nguyen-Van-Tam, J S Emergency (999) calls to the ambulance service that do not result in the patient being transported to hospital: an epidemiological study. Emerg Med J (2002), 19: 449 52. 10. Coid, D R Measurement for management: report of a pilot project to quantify ambulance misuse for managers of a Fife hospital. Health Serv Manage Res (1989), 2: 213 6. 11. Clark, M J, and FitzGerald, G Older people s use of ambulance services: a population based analysis. J Accid Emerg Med (1999), 16: 108 11. 12. Wilson, S, Edwards, S, and Cooke, M W Inappropriate ambulance use is a retrospective diagnosis. J Accid Emerg Med (1999), 16: 75. 13. Hauswald, M Can paramedics safely decide which patients do not need ambulance transport or emergency department care?. Prehosp Emerg Care (2002), 6: 383 6. 14. Kost, S, and Arruda, J Appropriateness of ambulance transport to a suburban pediatric emergency department. Prehosp Emerg Care (1999), 3: 187 90. 15. Canto, J G, Zalenski, R J, Ornato, J P, et al. Use of emergency medical services in acute myocardial infarction and subsequent quality of care: observations from the National Registry of Myocardial Infarction. Circulation (2002), 106: 3018 23. 16. Herlitz, J, Hjalmarson, A, Holmberg, S, Richterova, A, and Wennerblom, B Mortality and morbidity in suspected acute myocardial infarction in relation to ambulance transport. Eur Heart J (1987), 8: 503 9. 17. World Health Organization ICD-10: International Statistical Classification of Diseases and Related Health Problems,, 10th revision; World Health Organization: Geneva, Switzerland, (1992-1994). 18. Jelinek, G A, and Little, M Inter-rater reliability of the national triage scale over 11,500 simulated occasions of triage. Emerg Med (1996), 8: 226 30. 19. Altman, D G Practical Statistics for Medical Research Chapman Hall: London, (1994): 252. 20. Foundation of Clinical Research: Applications to Practice,, 2nd ed.; Portney, L G, and Watkins, M P, Eds.; Prentice Hall: Upper Saddle River, NJ, (2000): 597 602. 21. Applied Multiple Regression/ Correlation Analysis for the Behavioral Sciences,, 2nd ed.; Cohen, J, and Cohen, P, Eds.; Lawrence Erlbaum Associates: NJ, (1983): 79 98. 22. Afilalo, J, Marinovich, A, Afilalo, M, et al. Impact of ambulance transportation on resource use in the emergency department. Acad Emerg Med (2004), 11: 312 5. 23. Stathers, G M, Delpech, V, and Raftos, J R Factors influencing the presentation and care of elderly people in the emergency department. Med J Aust (1992), 156: 197 200.